vread
ai serviceOctober 2025 – present
A service that turns a YouTube link into text made for reading. Not a summary: the original, read at your own pace.
· foreign-language videos show each line of the original beside its Korean translation; Korean videos show the original as is · highlight by segment, hide sponsor reads, tap to jump the video to that point · a reading mode that moves at your pace and a listening mode that follows playback
· adjacent highlights merge into a single excerpt, each with its own note · every highlight is collected into a library automatically · export a single card holding the thumbnail and the quote
In the AI era, we didn't like taking in content through summaries alone. We wanted the whole thing, but videos are tied to their timeline, and many are in English. So we decided to read the original behind the video. vread turns a YouTube link into something you read, not just watch. The first version was a mini tool running on your own API key, built for ourselves. We used it every day, but it fell short as a product: setting up a key was a barrier for non-engineers, performance couldn't keep up with long videos, and the reading experience wasn't polished enough to recommend. So now we're rebuilding the tool we used alone into a service we can hand to others.
Built on tools of our own, walking the process again from planning. · planning: vibe ux skills carried it from idea through benchmark, brand voice, and user flows into a PRD; even the voice, "a quiet librarian who keeps the original," is set down in writing · mockup first: thirteen screens specced up front, then a frontend running on mock data; once the mockups were done, the PRD was rewritten with the implemented code as the source of truth · design: codexy design system (CDS) as the baseline for tokens and components · accessibility: graduated from bring-your-own-key to a credit-based model, with key registration kept as an option · stack: Next.js, Tailwind, Supabase